Kummer structures on a \(K3\) surface: An old question of T. Shioda (Q1430450)

The authors use the following notations and conventions: All varieties are smooth and projective over the complex numbers; \(X\) is a \(K3\) surface; \(A\), \(B\) are abelian surfaces; \(\widehat A\) denotes the Picard torus Pic\(^0(A)\); \(\rho(A)\) is the Picard number of \(A\); \(D(A)\) is the bounded derived category of coherent sheaves on \(A\); FM\((A)\) is the set of Fourier-Mukai partners of \(A\), i.e. all abelian surfaces \(B\) such that we have an equivalence of triangulated categories \(D(A) \cong D(B)\); Km\((A)\) denotes the Kummer surface of \(A\); \({\mathcal K}(X)\) the set of Kummer structures of the \(K3\) surface \(X\), i.e. all abelian surfaces \(A\) with Km\((A) \cong X\). This article mainly deals with the following question which originates in an old question of Tetsuji Shioda: Does Km\((A) \cong \)Km\((B)\) imply \(A \cong B\) or \(A \cong \hat B\)? The authors are able to answer this question. Their main theorem is the following result: Theorem 0.1. If \(X= \)Km\((A)\), then we have: (1) \({\mathcal K}(X)= \)FM\((A)\). Hence the set \({\mathcal K}(X)\) is finite and contains \(\widehat A\). (2) If \(\rho(A)=3\) and det(NS\((A)\)) is square free, then \({\mathcal K}(X)= \{ A, \widehat A \}\). If \(\rho(A)=4\), then \({\mathcal K}(X)= \{ A, \widehat A \}\). (3) For any natural number \(N\), there are abelian surfaces \(A_i\) (\(1 \leq i \leq N\)) such that Km\((A_i) \cong \text{}\)Km\((A_j)\) and \(\text{NS}(A_i) \not\cong \text{NS}(A_j)\) for all \(1 \leq i < j \leq N\). In an appendix, the authors show that the torus \(A = {\mathbb C}^2 / L\) for the general lattice \(L \subset {\mathbb C}^2\) satisfies NS\((A)=0\). They conclude (in contrast to the projective case) that \(A \not\cong \widehat A\) and \({\mathcal K}(\)Km\((A) ) = \{ A \}\). The proofs use facts about derived categories of \(K3\) and abelian surfaces, Fourier-Mukai transforms, Hodge isometries, Hodge structures and computations with lattices.